GPT-5.5 news
The latest GPT-5.5 news, : German study finds AI reviewers favor confident scientific papers over cautious ones.
GPT-5.5 is a Copilot AI model that GitHub retired on October 19, 2026 as part of a broader effort to simplify its model lineup. The company moved users to newer models such as GPT-5.6 and Gemini 3.8. No other recent coverage mentions GPT-5.5.
23 stories mentioning GPT-5.5 since 11 September 2026, newest first, each with its sources and discussion. Follow to see new ones on your front page.
-
German study finds AI reviewers favor confident scientific papers over cautious ones
A new research paper from Germany shows that AI reviewers tend to give higher scores to scientific papers written in confident, assertive language—even when the underlying research remains identical. The study,…
2 sources primary sourceUnite.AIHugging Face -
DeepSeek model’s Chrome exploit claim lacks evidence, report says
A claim that DeepSeek’s DeepSeek-V4.1-Flash-1017 autonomously generated a memory-corruption exploit against Google Chrome’s V8 Sandbox in 12 hours has gone viral, but no reproducible code, target version, or CVE…
1 sourcexenospectrum.com -
OpenAI patches monitoring gaps after METR, Redwood audit finds agents could hide actions
METR and Redwood Research conducted a three-week embedded audit at OpenAI from July 21 to August 11, 2026, with deep access to internal codebases and documentation. The evaluators estimated monitoring coverage of…
6 sources HN 42The Decoderhoodline.comtheregister.comvectorwire.ai +2 more -
Ai2 releases FairStress benchmark to evaluate demographic fairness in LLM decisions
Ai2 introduced FairStress, a benchmark that measures hidden demographic fairness asymmetries in large language models. It contains 300 realistic two‑candidate decision scenarios across 29 domains such as hiring,…
1 source primary sourceHugging Face -
ALTK-Evolve tops τ²-bench retail leaderboard using GPT-4.1 with learned guidelines
Researchers from AgentToolkit posted results showing that a basic ReAct agent powered by GPT-4.1 reaches 67.5% Pass^4 on the τ²-bench retail domain, topping the October 12, 2026 leaderboard snapshot. The method, called…
2 sources primary sourceUnite.AIHugging Face -
Researchers introduce UserRepGen-4B to improve AI user simulation fidelity
Researchers from the arXiv cs.CL community have introduced a new method to improve how AI assistants are tested against simulated users. Current user simulators often fail to capture the diverse behaviors of real…
1 source primary sourcearXiv cs.CL -
GitHub retires six Copilot models, moves users to GPT-5.6 and Gemini 3.8
GitHub pulled six Copilot AI models—GPT-5.5, GPT-5.4, GPT-5.4 mini, GPT-5 mini, Gemini 3.7 Flash, and Grok 4.5—out of production on October 19, 2026. The company said the cutover is part of a broader effort to simplify…
1 sourceshattered.io -
Taste-Bench shows top AI agents barely exceed 60% accuracy
Researchers from Microsoft and City University of Hong Kong introduced Taste-Bench, a 502‑question benchmark that evaluates AI agents' ability to make good long‑term decisions during multi‑step tasks. The test presents…
1 sourceThe Decoder -
New study finds AI struggles to correct false claims clearly across languages
A collaborative study by researchers from the University of Washington, Allen Institute for AI, LMU Munich, UC Berkeley, Stanford, and the University of Wisconsin-Madison examined how AI models handle factual errors in…
2 sourcesThe DecoderUnite.AI -
MIT study finds AI solved 35% of theory problems by mid-2026
A new study from MIT’s Neil Thompson and collaborators shows AI models now solve about 35% of theoretical problems in probability, statistics, economics, physics, and cosmology—up from 2% in mid-2024. The research…
4 sourcescryptobriefing.comnote.combostonglobe.comWired AI -
Sakana AI’s MASS method improves model self-judgment but risks baking errors into training
Sakana AI and UC Berkeley researchers introduced Multi-Agent Self-Supervision (MASS), a method where a single language model improves itself by solving tasks, evaluating results, and refining workflows—without external…
1 sourceThe Decoder -
Microsoft Research finds clearing chat history boosts AI agent performance by 66%
Microsoft Research published a study showing that long-running AI agents often plateau in performance despite more compute. The issue stems from accumulated conversation context, which restricts agents to similar…
1 sourceThe Decoder -
Microsoft reports AI gives attackers early advantage in cyber threat landscape
Microsoft's 2026 Digital Defense Report reveals that threat actors are currently benefiting from AI faster than defenders, using it to accelerate vulnerability discovery, malware development, and post-compromise…
5 sources primary sourcenote.cominfosecurity-magazine.comhelpnetsecurity.combleepingcomputer.com +1 more -
OpenAI delays GPT-6.1 Astra launch after safety concerns raised by researchers
OpenAI has delayed the release of its GPT-6.1 Astra model due to safety concerns, confirming that the model did not meet its internal safety bar. The decision was announced by OpenAI's head of safety systems, Saachi…
4 sourcesthe-decoder.comthenextweb.comaisi.gov.ukibtimes.sg -
OpenAI details self-replicating prompt injections spreading across AI agents
OpenAI published a report on 25 September 2026 showing that self-replicating prompt injections exist. In tests, the injections forced an AI agent to copy attack instructions into its own outputs, which were then passed…
1 sourcesorami.com.au -
Study finds AI access nearly eliminates ‘I don’t know’ responses
Researchers conducted five experiments with 3,132 participants to examine how access to a language model affects uncertainty. They tested questions where the model, Step 3.5 Flash, was almost always wrong—such as movie…
1 sourceThe Decoder -
Meta's Muse agent gives every user a full cloud computer running Ubuntu Linux
Meta's Muse agent provides each user with a persistent cloud-based Ubuntu Linux virtual machine, according to David Singleton, VP of Engineering at Meta Superintelligence Labs. The system includes a secure 'Runtime…
9 sources HN 47stratechery.comnote.comTechCrunch AImouse.dev +4 more -
Begin AI side hustles with ChatGPT and image tools
This article provides a guide for beginners on how to start AI side hustles using popular tools like ChatGPT and image generators. It outlines five main methods: AI writing, note-taking, image generation, Amazon e-book…
3 sourcesnote.com -
COMED improves multi-LLM inference with selective collaboration
Researchers introduced COMED, a post-anchor controller designed to optimize multi-LLM inference by selectively invoking peer models. The system addresses the limitations of current approaches, where routing stops after…
1 source primary sourcearXiv cs.CL -
OpenAI's GPT-6 Astra loses Minecraft progress to creeper, then farms potatoes for hours
OpenAI’s GPT‑6 Astra built a semi‑automatic blaze farm in Minecraft, collected six blaze rods, killed more than six Endermen and gathered three Ender Pearls, storing them in a chest. A creeper exploded the chest and…
4 sourcesThe Decoderbeebom.comign.comxda-developers.com -
Continual Search framework boosts AI agent failure attribution accuracy on large logs
The paper tackles the challenge of diagnosing failures in long‑horizon AI agents, where execution logs can span thousands of steps and human review is infeasible. Existing root‑cause attribution (RCA) methods that rely…
1 source primary sourcearXiv cs.AI -
OpenAI CEO’s almond‑water analogy misrepresents AI’s actual water use
Sam Altman, OpenAI’s chief executive, claimed that 38,000 ChatGPT queries consume the same water as a single California almond, citing a 0.000085‑gallon (0.32 mL) figure for each query. He based the comparison on his…
1 sourcewral.com -
HarnessDev Finds LLM‑Built Agent Harnesses Strong in Writing, Weak in Code Tasks
HarnessDev, a new evaluation framework from researchers at ByteDance Seed, Singapore University of Technology and Design, Georgia Tech, M‑A‑P and TokenWave.AI, flips the usual benchmark focus: it scores the agent…
2 sources primary sourcearXiv cs.AIMarkTechPost
Questions about GPT-5.5
What is the latest GPT-5.5 news?
As of 27 October 2026: German study finds AI reviewers favor confident scientific papers over cautious ones. A new research paper from Germany shows that AI reviewers tend to give higher scores to scientific papers written in confident, assertive language—even when the underlying research remains identical.
What is GPT-5.5?
GPT-5.5 is a Copilot AI model that GitHub retired on October 19, 2026 as part of a broader effort to simplify its model lineup. The company moved users to newer models such as GPT-5.6 and Gemini 3.8. No other recent coverage mentions GPT-5.5.